AltAnalyze is a multi-functional and easy-to-use software package for automated single-cell and bulk gene and splicing analyses. Easy-to-use precompiled graphical user-interface versions available from our website.
Background
In version 1.16 beta of AltAnalyze, we introduced the ASPIRE p-value as the
default metric for assessing reproducibility of ASPIRE scores for reciprocal
junction analysis of HJAY, MJAY and AltMouse arrays. Previously, the
appropriate recommended metric was the permutation p-value (still optional). An
issue with the permutation p-value is that it's power is dependent on the
number of possible permutation that result from the number of replicates.
The ASPIRE p-value was designed as an optional replacement for the permutation
p-value. This algorithm calculates and stores an ASPIRE scores for each
experimental sample for the two reciprocal probesets (normalized baseline
expression and normalized experimental expression), versus the mean normalized
baseline expression for all baseline samples (normalized baseline sample values
treated as experimental in the ASPIRE equation). These p-values are stored in
two different lists (baseline replicates versus baseline average and
experimental replicates versus baseline average). A one-way ANOVA p-value is
calculated between these two lists to derive the ASPIRE p-value.
Updates
While initially, this p-value seemed appropriate replacement, the baseline
replicate comparisons appear too variable when a small number of replicates is
present. Thus, we changed this metric as follows:
Normalized baseline replicate values are now compared to the mean normalized
experimental replicate values. Here, the baseline replicates are treated as
experimental and the experimental mean is treated as baseline in the ASPIRE
equation.
As before, experimental replicate values are compared to the baseline mean and
the two resulting replicate ASPIRE score lists are compared by one-way ANOVA.
Preliminary analysis of RNASeq data (version 2.0) where only two replicates are
present seems to indicate this is a reasonable measure of variability.
Nonetheless, we recommend using the FDR adjustment of this p-value, present in
the results file, as an additional filter.
Original issue reported on code.google.com by nsalomo...@gmail.com on 5 Feb 2011 at 6:52
Original issue reported on code.google.com by
nsalomo...@gmail.comon 5 Feb 2011 at 6:52